Wavelet Transform-based Automatic Classification Of EEG Signals In Mice Epilepsy Model

Epilepsy(EP)is a chronic neurological disorder in which sudden and recurrent brain cell populations in the brain undergo paroxysmal abnormal super-synchronous electrical activity,resulting in transient brain dysfunction.In order to further study the pathogenesis of epilepsy,a large number of laboratory studies often rely on animal models of epilepsy.The automatic classification of epileptic electroencephalogram(Epileptic EEG)is very important for the potential of mice models in epilepsy research.The related automatic classification algorithm of epilepsy EEG can also inspire the automatic classification algorithm of clinical epilepsy EEG.Helping to reduce the workload of medical workers and reduce the negative impact on patients.It is of great value and clinical significance in further research on the automatic detection of epileptic EEG signals.In order to study the automatic detection and classification of EEG in ani,mal models of epilepsy,a pentylenetetrazole(PTZ)induced acute epilepsy in mice was selected to record the administration time of epileptic agents,behavior,normal EEG and epileptic EEG in mice.Compared with epileptic diseases caused by human brain lesions,acute epileptic seizure symptoms induced by pentylenetetrazole drugs,electroencephalogram,sample epilepsy discharge and other aspects are generally similar,and the model of acute epilepsy induced by pentylenetetrazole in mice has been very mature,which is widely recognized in the world in the field of acute epileptics-related research,and is also a more idealized animal model.Based on the non-stationary(burst,amplitude anomaly)and frequency distribution characteristics of epilepsy EEG,this paper uses wavelet transform to decompose specific EEG components and extract features,and learn these features with support vector machine to realize mice epilepsy.Automatic classification of EEG.Firstly,the collected brain EEG original signals were segmented and preprocessed according to the epileptic administration time and mice behavior Secondly,the discrete stationary wavelet transform(SWT)was performed on the segmental EEG signals to obtain wavelets with different frequency sub-bands.Coefficient,select the wavelet coefficient in the frequency range of the epileptic EEG characteristic wave(spike wave,sharp wave,slow wave),and inverse wavelet transform to reconstruct the epileptic EEG characteristic wave band;The correlation linear features(maximum and standard deviation)and nonlinear features(sample entropy)were extracted from the pre-processed EEG signals,the related EEG wavelet coefficients and the reconstructed EEG;Finally,the support vector machine method was used to realize automatic classification of epilepsy EEG signals in mice.The experimental results show that the classification sensitivities based on the maximum,standard deviation and sample entropy of mouse EEG signals are 53.67%,62.00%and 48.33%,respectively,and the specificities are 87.67%,95.67%and 95.67%,respectively.The rates were 70.70%,80.00%and 72.00%respectively.After the characteristic fusion based on the maximum value,standard deviation and sample entropy of the relevant wavelet transform coefficients,the classification sensitivity of mouse epilepsy EEG reached 99.67%and the specificity was 100.00%,The correct rate reached 99.80%.On the other hand,after the feature fusion of the maximum value,standard deviation and sample entropy of the wavelet coefficients reconstructed by the wavelet coefficients in the frequency band of the epileptic wave,the classification sensitivity reached 99.00%.For 100.00%,the correct rate has also reached 99.50%.These results indicate that the classification algorithm based on wavelet decomposition coefficient or feature fusion of epileptic wave frequency band is significantly improved compared with the single feature learning algorithm based on signal itself,which can effectively realize the automatic detection and classification of mouse epilepsy EEG signals.